Remaining Useful Life Prediction Of Key Equipment Multiple Model Interaction

With the continuous development of science and the advance of technology in recent years,the level of industrialization is gradually increasing,and the operating environment of industrial equipment has become increasingly complex.The performance of the equipment during its operation will inevitably degrade,which will directly cause equipment failure.Once the equipment failure causes major accidents,it will face significant human and financial loss.At present,the repair method has been developed from the original passive repair to repair according to the plan or according to the situation.Compared with traditional passive maintenance,condition based maintenance has stronger pertinence,better maintenance effect and saves more manpower and financial resources.Prognostics and Health Management(PHM),as an important method to implement condition based maintenance,has been widely used in various fields.This technology consists of two parts:prognostics and health management.In practical application,it is very important to estimate and forecast the current and prospective state of equipment operation.And the key process is to predict the remaining useful life(RUL).Therefore,this paper concentrate on the equipments’ RUL prediction and carries out the following researches:(1)Aiming at the life prediction problem of key equipment with nonlinear degradation process,the degradation process was firstly analyzed and the degradation model was established.The model parameters were estimated to get the accurate state model,and the Extended Kalman filter and Particle Filter were used to predict the RUL of the key equipment.Lithium-ion battery data sets of CALCE and NASA were used for simulation verification,and the life prediction results were obtained.From the result,the relative errors were all less than 10%,showing a good prediction effect.(2)In view of the multi-stage degradation characteristics in the process of equipment degradation,it is difficult to accurately characterize the degradation process with a single model.A RUL prediction method based on multi-stage degradation process modeling is proposed in this paper.According to the degradation characteristics,the constant velocity degradation model and the constant acceleration degradation model were established.The interactive multi-model algorithm was used to predict the life of the simulation degradation process.By comparing the prediction results with those of a single model,it is proved that the method described in this paper has a better prediction effect.(3)According to single characteristic of the single model prediction results in the traditional life prediction of nonlinear degradation process,weak generalization ability and so on,the life prediction method for nonlinear degradation process based on multi-model interaction was proposed in this paper.The model of complex degradation process is modeled by multi-model to represent the transformation of degradation mode.According to the nonlinear non-gaussian characteristics of the degradation process,the Interacting Multiple Model Particle Filter was used in the life prediction.The simulation results are verified on the lithium battery data set,and the prediction results are compared with the single model prediction results.The results show that the life predicted method which is proposed in this paper is closer to the true remaining useful life,and the prediction error is smaller.